layer_norm 无法安装

[running-frog]

Warning: import flash_attn rms_norm fail, please install FlashAttention layer_norm to get higher efficiency https://github.com/Dao-AILab/flash-attention/tree/main/csrc/layer_norm

看了说明，说只适配A100，RTX3080怎么办？

[jklj077]

Flash Attention里有个新issue提了这事儿。应该是只在A100上做了测试，RTX 3080的按理说应该也行，架构一致。看Compute Capability的话¹，RTX 3080是8.6，A100是8.0，两者差异²可能会对kernel的实际效率有影响。

建议不装Flash Attention看看速度如何，如果满足需求的话，不是必需装它。README里也有开不开Flash Attention的实测效率数据（卡型确实不一样），可以参考看看。

Footnotes

1. https://developer.nvidia.com/cuda-gpus ↩

2. https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#compute-capability-8-x ↩